CAJ | 학술논문

根据粒子群算法求解多目标问题的特点，个体极值和全局极值的选择不同会对实验结果产生很大影响。目前普遍的选择方法仅仅根据简单的支配关系，但是会存在两个解之间没有支配关系而导致不去更新个体最优值(PB)和全局最优值(GB)，这样会导致更好的个体极值和全局极值的遗漏从而降低收敛时间。文中提出一种新的个体极值和全局极值的选择策略。使用这种策略，可以加快收敛，提高准确性，防止非劣解的遗漏。通过几个测试函数的实验仿真，所得解集的分步性和多样性都有显著的提高。
근거입자군산법구해다목표문제적특점，개체겁치화전국겁치적선택불동회대실험결과산생흔대영향。목전보편적선택방법부부근거간단적지배관계，단시회존재량개해지간몰유지배관계이도치불거경신개체최우치(PB)화전국최우치(GB)，저양회도치경호적개체겁치화전국겁치적유루종이강저수렴시간。문중제출일충신적개체겁치화전국겁치적선택책략。사용저충책략，가이가쾌수렴，제고준학성，방지비렬해적유루。통과궤개측시함수적실험방진，소득해집적분보성화다양성도유현저적제고。
According to the characteristics of particle swarm optimization algorithm for solving multi-objective problems,the choice of personal best and global best will affect the result greatly. The current selection method is only based on their dominant relationship,but if there is no dominant relationship between two solutions,PB and GB will not be updated. This will miss the better PB and GB and ex-tend the convergence time. A new selection strategy for personal best and global best is presented. Using this strategy can accelerate con-vergence,improve accuracy,avoid non-dominated solution discard. The performance of this strategy is evaluated on several test function. The results show that the diversity and the distribution of the non-dominated solution is highly raised compared with other PSO algo-rithm.